1. 1 Typing in Information Systems Nick Rossiter, Michael Heather Computer Science and Digital Technologies Northumbria University, Newcastle NE1 8ST, UK Nick.rossiter1@btinternet.com http://nickrossiter.org.uk/process/index.html http://computing.unn.ac.uk/staff/cgnr1/index.htm GRAPH OPERADS LOGIC AGMP X GOL Tallinn 25-28 June 2013

2. 2 Typing Most important property in computing science Defines for an attribute: – Permissible values – Permissible operations A type is a category of any complexity, including a discrete category

3. 3 Simplest Typing in Category Theory (a) A category with a terminal object (b) The opposite category with typing arrows Cartesian Closed (potentially) BB op or 1 B Identity Functor (potentially) 1 b : b → 1 B 1

4. 4 Nature of Typing Where does typing come from? Typing resides in the system itself. Typing is of the nature of the system and forms part of the Universe. Typing must therefore reside in nature and arise from relationships in nature. The Universe consists of entities related one to the other. Thus each entity affects every other.

5. 5 Existence and Cartesian Closure Existence is not just a first order effect but needs an inherent higher order formalism. The relationship between any pair of entities depends on every possible path between them. In category theory this is the property of cartesian closure found in the highest structure possible -- the identity natural transformation designated as the topos. However if every entity is related to every other it follows that the relationship is both ways but not just a simple inverse relationship as appears from the laws of physics.

6. 6 Duality ● A category C of objects and arrows between the objects will have a dual C op with arrows reversed. ● The whole universal structure of both-ways relationships will then be represented by the product C op x C. ● This gives rise to the principle of duality throughout the Universe. ● C op x C is cartesian closed, with products, terminal object and exponentials

7. 7 Ubiquity of Duality ● Duality is a common enough concept in mathematics, philosophy and most of the sciences with some renowned examples like the mind-body duality. - It also appears in other versions of contrast as between the dynamic and the static and between global and local. ● To capture the full effect and subtleties of opposing views and relationships a single view of the duality is needed as a process (e.g. a monad).

8. 8 Duality and Variance ● Duality is not a closed Boolean view. - Rather it encapsulates opposite orderings within a single (functorial) concept of variancy. - These may be conveniently labelled covariant and contravariant but only relative one to the other and not as absolute descriptions. ● Systems theory is a case in point where these different views need to be integrated. - Thus for object-oriented computing systems, covariance is assumed for specialisation (looking forward) and contravariance for generalisation (looking back). ● The natural categories of process as advanced by Whitehead encompass this contravariancy found in reality.

9. 9 Covariant and Contravariant Functors Opposite Covariant F Contravariant F-bar

10. 10 Contravariancy Highlighted by Lawvere in 1969 as basic property in the intension-extension relationship Governing data values in the context of their name and type Basic property of universe Lawvere defined the relationship between intension and extension in terms of adjoint functors with contravariant mapping – Used concept of hyperdoctrine – Some 'translation' needed for applied science

11. 11 Why Contravariant? The extension is of the form: value → nameN:1 The intension is of the form: name → typeN:1 If these arrows were reversed, they would not be determinations (functions) so can reject such forms: name → value 1:N type → name 1:N

12. 12 Turn around one arrow But the common attribute in extension and intension – name – is codomain in extension and domain in intension. So cannot do simple covariant mapping of one to the other. Need to turn around the arrow in the intension name → typetype → name And map this onto value → name in extension So that value is related to type in the context of a common name

13. 13 Contravariancy Technicalities source arrow Apply functor F Turn source round Map onto target Square must commute composition

14. 14 Ultimate Contravariancy A three-level structure is sufficient to provide complete closure with internal contravariant logic providing a generalisation of negation. – Further levels are redundant Contravariancy across levels provides more sophisticated reversals such as reverse engineering. The ultimate contravariancy is to be found in the universal adjointness - between any pair of functors contravariant one to the other - to provide both the quantitative and qualitative semantics of intension-extension logic.

15. 15 Worked Example of Three-level Architecture Choices for realisation in formal terms Informal look at structures and relationships Outline in informal categories Two-way mappings as adjunctions Examples of Contravariancy

16. 16 MetaMeta Policy Meta Organize Classify Instantiate Concepts Constructs Schema Types Named Data Values Downward arrows are intension-extension pairs Figure 1: Informal requirements for Information System Architecture

17. 17 Formalising the Architecture Requirements: –mappings within levels and across levels –bidirectional mappings –closure at top level –open-ended logic –relationships (product and coproduct) Choice: Category theory as used in mathematics as a workspace for relating different constructions

18. 18 Figure 2: Interpretation of Levels as Natural Schema in General Terms blue – category, red - functor, green - natural transformation

19. 19 black - objects Figure 4: Defining the Three Levels with Contravariant Functors and Intension-Extension (I-E) Pairs P -| P' O -| O' I -| I'

20. 20 Figure 3: Example for Comparison of Mappings in two Systems Categories: CPT concepts, CST constructs, SCH schema, DAT data, Functors: P policy, O org, I instance, Natural transformations: , ,  (Organisational interoperability)

21. 21 Figure 5: Examples of Levels in the Three Level Architecture Cross-over arrows indicate contravariant mapping

22. 22 Figure 6: Composition of Adjoints is Natural If functors are adjoint, there is a unique relationship between them (a natural bijection). Can write for instance IOP -| P'O'I' and OP -| P'O'

23. 23 Adjointness Probably the most important feature of category theory Deals with relative ordering in duality between functors For dual functors L and R written (if holds): L --| R meaning L is left adjoint to R and R is right adjoint to L Unique solution to dual relationship between functors Full expression is a 4-tuple

24. 24 Adjointness between Functors F and G mapping Categories L and R F ┤ G

25. 25 Composition Triangles in Detail a) unit of adjunction  ; b) co-unit of adjunction  If isomorphism, diagram collapses with η = 1 and ε = 0 1 L = GF(L); FG(R) = 1 R